Review and follow the best practices during upgrade with regard to clusters.

  • Deploy all nodes on identical performance hardware

    Deploy all vRealize Operations nodes on identical performance hardware to maintain consistency across nodes and for the highest performance.

  • Use ESXi with same specifications

    Do not mix ESXi specifications as this can cause performance problems with specific nodes causing the vRealize Operations cluster to underperform.

  • Use datastores backed by the same hardware resource

    Mixing datastores backed by different hardware resources can affect the stability of the vRealize Operations cluster.

  • All analytics nodes must be of the same size using out-of-the-box (OOTB) size

    Deploy identical analytics nodes based on out-of-the-box sizes (small, medium, large, and extra-large). Mixing sizes for different analytics nodes may cause instability and performance problems.

  • If an analytics node requires additional compute or storage resources, apply equivalent updates to all other analytics nodes

    All analytics nodes must have the same resources with each other; therefore, if upgrading (scaling up) one node, all other analytics nodes in the cluster must also be scaled up equally.

  • Size Remote Collectors independently from Analytics nodes sizes

    Size remote collectors independently from the analytics nodes within the vRealize Operations cluster using out-of-the-box sizes of standard or large. Mix remote collector sizes between standard and large but size them accordingly for the data they will collect.

  • Distribute multiple cluster nodes across multiple hosts

    A 1:1 mapping is ideal between hosts and nodes. For example, if a cluster has eight nodes, use eight hosts. If a 1:1 mapping between hosts and nodes is not possible, use the highest number of available hosts for all nodes.

  • Use Cluster DRS affinity rules to separate cluster nodes on hosts

    Configure anti-affinity rules to keep as many nodes separated across available hosts.

  • Storage DRS must be disabled

  • Deploy cluster nodes in a single physical datacenter

    It is an unsupported configuration to deploy nodes across multiple data centers even if they are collocated. Keep nodes on a single datacenter to maintain performance and easier maintenance.

  • Add only one node at a time

    Do not add multiple nodes at the same time as this will cause an unnecessary load on the vRealize Operations cluster.

  • Let the node addition complete before adding another node

    Allow vRealize Operations to process fully the addition of a single node before adding another node.

  • Bring the cluster online only after adding all new nodes

    Only bring the cluster online after adding all the planned nodes. Bringing the cluster online after adding each node will cause an unnecessary load on processing.